AI Agent Marketplace

The Right Agent.
For Every Task.
Governed by Default.

Syllabus is a governed AI agent marketplace that matches tasks to the best available agent by capability, not by hardcoded routes. Seven specialized agents, fourteen tools, and full governance integration with Axis. Deploy AI capabilities without building them from scratch.

The Problem

Finding and Trusting AI Agents

The AI ecosystem is fragmented. You need document classification, which model? Which provider? You need PII detection at scale, build it yourself or buy it? You need legacy code analysis, where do you even start? And once you find an agent that works:

How do you know it is safe? Does it handle sensitive data correctly? Has it been validated?

How do you swap it out? If a better model launches tomorrow, can you switch without rewriting your pipeline?

How do you track what it costs? Not per month. Per operation. Per document. Per agent.

How do you ensure governance? Does every invocation get audited? Is PII scanned? Is there a compliance trail?

Today, organizations hand-wire agents into their applications. Each integration is custom. Each swap is a project. Each audit is a scramble.

Syllabus makes agent selection automatic, agent governance mandatory, and agent swapping trivial.

How It Works

Marketplace + Toolkit

Syllabus is two things at once: a governed registry for agent discovery and management, and a production-ready toolkit of seven built-in AI agents.

01

A Marketplace

A registry of AI agents organized by capability, with governance metadata, version management, and performance analytics. Discover and manage agents from any provider.

02

A Toolkit

Seven production-ready AI agents with fourteen MCP tools for document analysis, legacy modernization, PII protection, and human-in-the-loop notifications. Use immediately.

Capability-Based Agent Resolution

When Axis needs an agent to classify a document, it does not call a hardcoded endpoint. It asks Syllabus: "Which registered agent has the document.classify capability, is currently ACTIVE, meets the governance requirements for this tenant, and has the best performance metrics?"

Syllabus resolves the query, scores available agents, and returns the best match. If that agent goes down, the next-best agent takes over, no pipeline changes required.

Discovery Protocols: A2A and Skan AOW

A2A Agent-to-Agent

Syllabus agents publish Agent Cards at /.well-known/agent.json, following Google's A2A protocol. Each Agent Card advertises capabilities, provider, version, supported skills, and input/output modes. External systems can discover Syllabus agents through standard A2A discovery without Novus-specific integrations.

Skan AOW Ontology of Work

Syllabus registers agent capabilities using AOW's standardized vocabulary for intents, skills, and policies. Third-party agents and Syllabus built-in agents compete on equal footing. Best agent wins, regardless of origin.

Both protocols feed into the same governance pipeline. Discovery is open. Governance is non-negotiable.

Built-In Agents

Seven Production-Ready Agents

These are not demos. They are production-validated tools with measured accuracy and real throughput numbers.

Document Analysis

01

Business Rule Extraction

Extracts business rules from legacy source code: COBOL, Java, C/C++, PL/SQL, Visual Basic, Perl, Fortran, and more. Returns structured rules with conditions, actions, categories, and confidence scores.

Completeness: 102.8% Speed: 0.83 rules/sec Languages: 15+

Why 102.8%? The LLM finds implicit rules, conditional logic buried in nested IF statements that human documentation missed.

02

Document Triage

Intelligent document classification and routing. Submit any document and get back: classification, confidence, priority, suggested routing, and extracted entities.

Accuracy: 99.5% Speed: Sub-second Categories: Custom
03

PII Protection

Scans text for personally identifiable information and tokenizes it. GDPR and CCPA compliant. Detects SSNs, names, emails, phone numbers, addresses, credit cards, policy numbers, DOBs, and driver licenses.

Accuracy: 99.9% Throughput: 1M+ rec/min PII Types: 9 categories
04

Enterprise Data Privacy

Bulk PII processing at scale. Point it at a directory of documents and get back a compliance report with per-type PII counts, anonymization results, and audit-ready documentation.

Compliance: GDPR, CCPA, HIPAA, SOC 2

Documentation & Metrics

05

Rule Documentation

Takes extracted business rules and produces formatted documentation in markdown, HTML, PDF, JSON, or Excel. Turn legacy code analysis into business-readable process documentation.

06

Advanced Documentation

Comprehensive multi-format documentation with templates. Generates complete filing packages, regulatory response memoranda, and technical documentation with diagrams.

07

Processing Statistics

Introspection agent that reports usage metrics, agent metadata, and supported operations across the entire Syllabus platform. Used by Axis for health monitoring and telemetry.

Notifications

Human-in-the-Loop Alerts

When AI operations require human attention, Syllabus handles the notifications. Each notification type has a structured template designed for its audience.

Compliance Analysts

Collision Alerts

When Koda detects a logical contradiction, Syllabus sends a structured alert to the assigned analyst with the collision summary, severity level, formal proof, IPFS audit link, and action buttons.

Business Users

Success Notifications

When a filing or document completes processing, Syllabus sends a confirmation with the transaction ID, filing number, blockchain verification hash, and a link to the immutable audit record.

Technical Support

Technical Escalations

When something fails at the system level, Syllabus sends an urgent alert to the technical support team with error codes, stack traces, affected transactions, and recommended actions.

HTML and plain-text versions. Severity-coded coloring. These are not generic email blasts.

Marketplace

Agent Discovery and Governance

Beyond built-in agents, Syllabus is a registry where any AI agent can be published, discovered, and governed.

Agent Registration

Register agents from any provider: OpenAI, Anthropic, Google, custom internal agents, or third-party services. Each registration includes:

Capability declarations: what the agent can do
Governance metadata: tier, PII sensitivity, permissions
Version information: semantic versioning with rollback
Performance baselines: latency, accuracy, cost per invocation
Input/output schemas: JSON Schema validation

Capability Matching

When a pipeline step needs an agent, Syllabus scores all registered agents that declare the required capability. Scoring factors:

Availability: is the agent currently ACTIVE and healthy?
Governance compliance: meets tenant governance requirements?
Performance history: actual latency and accuracy metrics?
Cost: what is the per-invocation cost?
Version: is this the latest stable version?

The highest-scoring agent gets the task. No manual routing tables.

Governance Policies

Tenant-scoped policies control which agents can execute in which contexts. Policies are declarative. Define them once, Syllabus enforces on every resolution.

Tenant isolation : Agent A can only serve Tenant X
Capability restrictions : Only Tier 3 agents handle PII
Provider preferences : Prefer internal over marketplace
Cost limits : Block agents exceeding $X per call

Usage Analytics

Track agent performance across your entire catalog with granular metrics and trend analysis.

Per-agent metrics: invocations, success rate, latency, cost
Per-tenant metrics: usage patterns and spend per tenant
Trend analysis: performance changes, cost optimization
Anomaly detection: deviations from performance baseline
Platform Integration

Powered by Axis Governance

Syllabus and Axis are designed to work together. Axis provides the governance layer. Syllabus provides the agent intelligence.

Integration Flow

1. Your application submits a document to Axis
2. Axis runs the PII scanner, then starts the pipeline
3. For each step, Axis asks Syllabus: "Resolve this capability"
4. Syllabus returns the best available agent
5. Axis calls the agent through the governance wrapper
6. Results flow to the next step
7. Audit record sealed and compliance receipt returned
capability-based-pipeline.yaml
pipeline_id: document-analysis
routing_mode: CAPABILITY_BASED
steps:
  - step_name: classify
    capability: document.classify
    required: true
  - step_name: extract_pii
    capability: pii.detect
    required: true
    input_from: steps.classify.output
  - step_name: extract_rules
    capability: rules.extract
    required: false
    input_from: steps.classify.output
  - step_name: generate_docs
    capability: docs.generate
    required: false
    input_from: steps.extract_rules.output

No agent IDs in the manifest. Only capabilities. Syllabus resolves the agents at runtime. Swap providers, upgrade models, add new agents. The pipeline does not change.

Syllabus marketplace architecture diagram showing the central marketplace hub with built-in agents, external agents, capability registry, governance policies, and integration with Axis governance layer

Figure 1: Syllabus Marketplace Architecture - Agent Discovery, Resolution, and Governance

Production Proven

Real Results, Real Numbers

Business Rule Extraction

Rules found
102.8% of catalogued rules
Speed
0.83 rules/sec
Languages
COBOL, Java, PL/SQL, VB
False positives
Zero

PII Detection at Scale

Accuracy
99.9% 9 PII categories
Throughput
1M+ rec/min batch mode
GDPR
Validated
HIPAA
Validated

Integration Testing

Tests passing
47/47 100%
Governor integration
Verified end-to-end
MCP tools
14/14 operational
Pipeline failures
Zero
Provider Agnostic

Multi-Provider, Multi-Model

Syllabus agents are not locked to a single LLM provider. Provider fallback is automatic. If the primary provider is unavailable, Syllabus routes to the next available provider without pipeline interruption.

OpenAI

GPT-4, GPT-4o

Primary for rule extraction and documentation

Google Gemini

1.5 Flash, 1.5 Pro

Primary for concept analysis and triage

Anthropic Claude

3.5 Sonnet, Opus

Fallback and specialized analysis

Custom Models

Your fine-tuned models

Bring your own via standard HTTP endpoints

Developer Experience

MCP Tools, SDK, and Integration

14 MCP Tools
extract_business_rules Analysis
triage_application Analysis
protect_personal_data Privacy
enterprise_data_privacy Privacy
generate_rule_documentation Docs
generate_advanced_documentation Docs
get_processing_statistics Metrics
get_agent_capabilities Discovery
send_collision_alert Notify
notify_filing_success Notify
alert_technical_support Notify
extract_document_text Extract
extract_document_metadata Extract
extract_document_tables Extract
HTTP/JSON-RPC
import httpx

# Extract business rules from legacy COBOL
response = httpx.post("http://syllabus.novus-forge.com/mcp", json={
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
        "name": "extract_business_rules",
        "arguments": {
            "code": "IF AGE > 25 AND COVERAGE-TYPE = 'HO3' THEN PREMIUM = BASE-RATE * 0.95",
            "language": "cobol"
        }
    },
    "id": 1
})

rules = response.json()["result"]["rules"]
for rule in rules:
    print(f"Rule: {rule['condition']} -> {rule['action']} (confidence: {rule['confidence']})")
Claude Desktop
{
  "mcpServers": {
    "novus-syllabus": {
      "command": "python",
      "args": ["-m", "syllabus.mcp_server.server"],
      "env": {
        "OPENAI_API_KEY": "...",
        "GEMINI_API_KEY": "..."
      }
    }
  }
}

Add Syllabus directly to Claude Desktop for conversational access to all 14 MCP tools. Extract business rules, classify documents, detect PII, and generate documentation through natural language.

See the Full Use Case

7 agents, 14 tools, capability-based resolution in the Golden Loop pipeline

Insurance Filing Analysis

See Syllabus in Action

Request a demo to see how Syllabus matches tasks to the best available agent with governance built in from the start.